17 January 2020 in IEEE Communications Surveys & Tutorials
Polar codes have attracted the attention of academia and industry alike in the past decade, such that the 5th generation wireless systems (5G) standardization process of the 3rd generation partnership project (3GPP) chose polar codes as a channel coding scheme. In this tutorial, we provide a description of the encoding process of polar codes adopted by the 5G standard. We illustrate the struggles of designing a family of polar codes able to satisfy the demands of 5G systems, with particular attention to rate flexibility and low decoding latency. The result of these efforts is an elaborate framework that applies novel coding techniques to provide a solid channel code for NR requirements.
17 February 2022 in IEEE Transactions on Circuits and Systems I: Regular Papers
Guessing Random Additive Noise Decoding (GRAND) is a universal decoding algorithm that has been recently proposed as a practical way to perform maximum likelihood decoding. It generates a sequence of possible error patterns and applies them to the received vector, checking if the result is a valid codeword. Ordered reliability bits GRAND (ORBGRAND) improves on GRAND by considering soft information received from the channel. Both GRAND and ORBGRAND have been implemented in hardware, focusing on average performance, sacrificing worst case throughput and latency. In this work, an improved pattern schedule for ORBGRAND is proposed. It provides $> 0.5$ dB gain over the standard schedule at a block error rate $\le 10^{-5}$ , and outperforms more complex GRAND flavors with a fraction of the complexity. The proposed schedule is used within a novel code-agnositic decoder architecture: the decoder guarantees fixed high throughput and low latency, making it attractive for latency-constrained applications. It outperforms the worst-case performance of decoders by orders of magnitude, and outperforms many best-case figures. Decoding a code of length 128, it achieves a throughput of 79.21 Gb/s with 58.49 ns latency, yielding better energy efficiency and comparable area efficiency with respect to the state of the art.
17 October 2017 in IEEE Transactions on Circuits and Systems I: Regular Papers
In the past few years, the demand for real-time hardware implementations of deep neural networks (DNNs), especially convolutional neural networks (CNNs), has dramatically increased, thanks to their excellent performance on a wide range of recognition and classification tasks. When considering real-time action recognition and video/image classification systems, latency is of paramount importance. Therefore, applications strive to maximize the accuracy while keeping the latency under a given application-specific maximum: in most cases, this threshold cannot exceed a few hundred milliseconds. Until now, the research on DNNs has mainly focused on achieving a better classification or recognition accuracy, whereas very few works in literature take in account the computational complexity of the model. In this paper, we propose an efficient computational method, which is inspired by a computational core of fully connected neural networks, to process convolutional layers of state-of-the-art deep CNNs within strict latency requirements. To this end, we implemented our method customized for VGG and VGG-based networks which have shown state-of-the-art performance on different classification/recognition data sets. The implementation results in 65-nm CMOS technology show that the proposed accelerator can process convolutional layers of VGGNet up to 9.5 times faster than state-of-the-art accelerators reported to-date while occupying 3.5 mm².
2 October 2018 in IEEE Transactions on Communications
The interest in polar codes has been increasing significantly since their adoption for use in the 5th generation wireless systems standard. Successive cancellation (SC) decoding algorithm has low implementation complexity, but yields mediocre error-correction performance at the code lengths of interest. SC-Flip algorithm improves the error-correction performance of SC by identifying possibly erroneous decisions made by SC and re-iterates after flipping one bit. It was recently shown that only a portion of bit-channels are most likely to be in error. In this work, we investigate the average log-likelihood ratio (LLR) values and their distribution related to the erroneous bitchannels, and develop the Thresholded SC-Flip (TSCF) decoding algorithm. We also replace the LLR selection and sorting of SCFlip with a comparator to reduce the implementation complexity. Simulation results demonstrate that for practical code lengths and a wide range of rates, TSCF shows negligible loss compared to the error-correction performance obtained when all singleerrors are corrected. At matching maximum iterations, TSCF has an error-correction performance gain of up to 0.45 dB compared with SC-Flip decoding. At matching error-correction performance, the computational complexity of TSCF is reduced by up to 40% on average, and requires up to 5× lower maximum number of iterations.
1 January 2020 in IEEE Transactions on Computers
Convolutional neural networks (CNNs) are a vital approach in machine learning. However, their high complexity and energy consumption make them challenging to embed in mobile applications at the edge requiring real-time processes such as smart phones. In order to meet the real-time constraint of edge devices, recently proposed custom hardware CNN accelerators have exploited parallel processing elements (PEs) to increase throughput. However, this straightforward parallelization of PEs and high memory bandwidth require high data movement, leading to large energy consumption. As a result, only a certain number of PEs can be instantiated when designing bandwidth-limited custom accelerators targeting edge devices. While most bandwidth-limited designs claim a peak performance of a few hundred giga operations per second, their average runtime performance is substantially lower than their roofline when applied to state-of-the-art CNNs such as AlexNet, VGGNet and ResNet, as a result of low resource utilization and arithmetic intensity. In this work, we propose a zero-activation-skipping convolutional accelerator (ZASCA) that avoids noncontributory multiplications with zero-valued activations. ZASCA employs a dataflow that minimizes the gap between its average and peak performances while maximizing its arithmetic intensity for both sparse and dense representations of activations, targeting the bandwidth-limited edge computing scenario. More precisely, ZASCA achieves a performance efficiency of up to 94% over a set of state-of-the-art CNNs for image classification with dense representation where the performance efficiency is the ratio between the average runtime performance and the peak performance. Using its zero-skipping feature, ZASCA can further improve the performance efficiency of the state-of-theart CNNs by up to 1.9× depending on the sparsity degree of activations. The implementation results in 65-nm TSMC CMOS technology show that, compared to the most energy-efficient accelerator, ZASCA can process convolutions from 5.5× to 17.5× faster, and is between 2.1× and 4.5× more energy efficient while occupying 2.1× less silicon area.
18 October 2017 in IEEE Journal on Emerging and Selected Topics in Circuits and Systems
Polar codes have gained a great amount of attention in the past few years, since they can provably achieve the capacity of a symmetric channel with a low-complexity encoding and decoding algorithm. As a result, polar codes have been selected as a coding scheme in the 5th generation wireless communication standard. Among different decoding schemes, successive-cancellation (SC) and SC list decoding yield good trade-off between error-correction performance and hardware implementation cost. However, both families of algorithms have large memory requirements. In this paper, we propose a set of novel techniques that aim at reducing the high-memory cost of SC-based decoders. These techniques are orthogonal to the specific decoder architecture considered, and can be applied on top of existing memory reduction techniques. We have designed and implemented different polar decoders on FPGA and also synthesized them in 65 nm TSMC CMOS technology to verify the effectiveness of the proposed memory reduction techniques. The benchmark decoders yield comparable or lower area occupation than the state of the art: the results show that the proposed methods can save up to 46% memory area occupation and 42% total area occupation compared with benchmark SC-based decoders.